Proteinases of the bone morphogenetic protein-1 family convert procollagen VII to mature anchoring fibril collagen.

Collagen VII is the major structural component of the anchoring fibrils at the dermal-epidermal junction in the skin. It is secreted by keratinocytes as a precursor, procollagen VII, and processed into mature collagen during polymerization of the anchoring fibrils. We show that bone morphogenetic protein-1 (BMP-1), which exhibits procollagen C-proteinase activity, cleaves the C-terminal propeptide from human procollagen VII. The cleavage occurs at the BMP-1 consensus cleavage site SYAA/DTAG within the NC-2 domain. Mammalian tolloid-like (mTLL)-1 and -2, two other proteases of the astacin enzyme family, were able to process procollagen VII at the same site in vitro. Immunohistochemical and genetic evidence supported the involvement of these enzymes in cleaving type VII procollagen in vivo. Both BMP-1 and mTLL-1 are expressed in the skin and in cultured cutaneous cells. A naturally occurring deletion in the human COL7A1 gene, 8523del14, which is associated with dystrophic epidermolysis bullosa and eliminates the BMP-1 consensus sequence, abolished processing of procollagen VII, and in mutant skin procollagen VII accumulated at the dermal-epidermal junction. On the other hand, deficiency of BMP-1 in the skin of knockout mouse embryos did not prevent processing of procollagen VII to mature collagen, suggesting that mTLL-1 and/or mTLL-2 can substitute for BMP-1 in the processing of procollagen VII in situ.

Post-translational proteolytic processing of proteins is emerging as a major regulatory mechanism in biology (1). A wide spectrum of extracellular matrix molecules is proteolytically cleaved to yield mature biosynthetic products and to release functionally important domains or biologically active fragments. Examples are the processing of procollagens (2-7), proteoglycans (8) or laminins (9,10) to mature molecules, the shedding of ectodomains of transmembrane components, e.g. collagen XVII 1 or syndecans (11), or the release of biologically active fragments, such as endostatin (12). Similarly, processing of cell adhesion molecules allows regulation of cell attachment and migration (13).
The dermal-epidermal junction is a highly specialized basement membrane zone, which ensures the adhesion of the epidermis and the dermis in the skin. It contains distinct multicomponent aggregates (i.e. hemidesmosomes, anchoring filaments, and anchoring fibrils), which interact with each other to provide a tight linkage of the skin layers, and to mediate cross-talk between the epidermal keratinocytes and dermal fibroblasts (26). Collagen VII is a major protein of this zone and the main structural component of the anchoring fibrils. Keratinocytes synthesize and secrete it into the extracellular space as a precursor, procollagen VII; during fibril polymerization, a C-terminal propeptide is proteolytically released (27)(28)(29). In normal human skin, the C-propeptide is completely removed and is not found at the dermal-epidermal junction. However, in dystrophic epidermolysis bullosa (DEB), a genetic disease caused by mutations of collagen VII (30,31), procollagen can be retained in the skin of the patient, implying that deficient C-propeptide processing is associated with functional abnormalities of the anchoring fibrils (32). The importance of the NC-2 domain for antiparallel dimer formation was recently demonstrated by Chen et al. (33). However, so far, the mechanisms of fibrillogenesis of the anchoring fibrils, including the propeptide cleavage site and the enzyme(s) involved, have remained elusive. Two lines of evidence point to BMP-1 as the candidate enzyme for the processing of procollagen VII. First, the NC-2 domain of procollagen VII contains a BMP-1 consensus sequence motif (8,34). Second, laminin 5, a binding ligand of collagen VII at the dermal-epidermal junction (35), is processed by BMP-1 (9,36).
The proteolytic cleavage of laminin 5 ␣3 and ␥2 chains is believed to regulate the interactions between the laminin molecule and its ligands, integrin ␣ 6 ␤ 4 , laminin 6, and collagen VII. The cleavage of the ␥2 chain drives deposition of laminin 5 into the extracellular matrix and sustains cell adhesion (10). In concert, malignant cells with high migratory activity do not process laminin 5 completely (37), and in a cylindroma tumor model, abnormal processing of both chains was associated with striking structural abnormalities of the basement membrane (38).
The above observations argue for the functional importance of proteolytic processing of the epidermal adhesion partners laminin 5 and collagen VII. Here we show that the same enzyme(s) is/are responsible for the cleavage of both counterparts and postulate that the cleavage of the keratinocyte-derived laminin 5 and procollagen VII by fibroblast-derived enzyme(s) presents a mechanism of regulation of the assembly of the dermal-epidermal junction and an intriguing manifestation of epithelial-mesenchymal interactions.

EXPERIMENTAL PROCEDURES
Cell Cultures-Primary human keratinocytes and HaCaT cells (kindly provided by Norbert E. Fusenig, DKFZ Heidelberg, Germany) were cultured in keratinocyte growth medium supplemented with 50 g/ml bovine pituitary extract and 5 ng/ml recombinant human epidermal growth factor (Invitrogen) as described (39). Primary human skin fibroblasts were grown from skin explants as described (39). Prior to protein analyses, the cells were transferred to a medium without supplements and were grown in the presence of 50 g/ml L-ascorbic acid for 48 h.
Protein Extraction-For extraction of procollagen VII, the cell layers were washed with cold phosphate-buffered saline and extracted as described previously (40). Extraction of proteins from normal or DEB human skin biopsies and normal or Bmp1 Ϫ/Ϫ mouse embryo skin (41) was performed as described (42).
Mutated Procollagen VII-Mutated procollagen VII was isolated from cultured keratinocytes of two DEB patients with the methods described above. COL7A1 mutation analysis was performed with PCR amplification of all 118 COL7A1 exons directly from genomic DNA of the patients, followed by heteroduplex analysis and dideoxynucleotide sequencing, as described previously (40). Both patients carried the deletion mutation 8523del14 on one allele of the COL7A1 gene (32,43). The mutation leads to in-frame skipping of exon 115, which encodes a sequence of 29 amino acids containing the BMP-1 cleavage site. On the other allele, patient 1 carried the known nonsense mutation A425G (44). Thus, patient 1 was functionally homozygous for the deletion, i.e. synthesized only homotrimers consisting of deleted pro␣(VII) chains. Patient 2 had a glycine substitution mutation G2009R on the other allele and was thus compound heterozygous for two different mutations (32).
SDS-PAGE and Immunoblotting-For SDS-PAGE analysis, proteins were precipitated with trichloroacetic acid in the presence of deoxycholic acid (45). Immunoblotting was carried out using standard techniques. The domain-specific collagen VII antibodies are as follows: NC1-F3 to the NC-1 domain, 3 VII-aff against the triple-helical domain (46), and NC2-7 and NC2-10 against the NC-2 domain of procollagen VII (32). The antibody NC2-7 recognizes epitopes within the propeptide, and therewith only procollagen VII but not mature collagen VII. The secondary antibody was peroxidase-conjugated anti-rabbit IgG (Sigma). The signals were detected with the Renaissance chemiluminescence substrate from PerkinElmer Life Sciences.
Production of Recombinant Truncated Procollagen VII-One g of total RNA from cultured human foreskin keratinocytes was reversetranscribed, and PCR was performed following the manufacturer's instructions (Herculase Enhanced DNA Polymerase, Stratagene, La Jolla, CA). The following fragments of human collagen VII cDNAs were generated and linked together by overlapping PCR: human collagen VII (NM_000094), nucleotides 3237-3965 and 8163-8945. The PCR product was subcloned (Rapid DNA Ligation Kit; Roche Diagnostics) into the episomal expression vector pCEP-Pu/BM40 (47) (a kind gift from Ernst Pöschl, University of Erlangen, Germany). For convenience, an octa-His tag followed by a stop codon was introduced at the 3Ј end, adjacent to the BamHI site of the vector. TOP 10 cells (Invitrogen) were transformed with the vector. Plasmids were isolated from the bacteria (Qiaprep, Qiagen, Hilden, Germany) and sequenced with gene-specific primers (Thermo Sequenase Radiolabeled Terminator Cycle Sequencing Kit; U. S. Biochemical Corp.). 293-EBNA cells (Invitrogen) were transfected (FuGENE, Roche Diagnostics) with the expression vector and selected after 2 days with puromycin (Sigma). Transfected 293-EBNA cells were subcloned, and the clones with the highest protein production were chosen for large scale production. The cells were cultured in the presence of puromycin and of 50 g/ml ascorbic acid for hydroxylation of prolyl and lysyl residues within the triple-helical domain. Two liters of supernatant from these cells were collected and supplemented with 0.5 mM Pefabloc (AEBSF, Merck). After ammonium sulfate precipitation (45% saturation), the precipitate was collected by centrifugation and then dialyzed against binding buffer (200 mM NaCl, 20 mM Tris-HCl, pH 8). The dialyzed protein was applied onto a nickelchelated Sepharose column (Amersham Biosciences AB) and eluted with binding buffer containing increasing concentrations of imidazole (10 -80 mM imidazole). The protein was dialyzed against phosphatebuffered saline, and the protein concentration was determined using the BCA protein assay reagent (Pierce). Rotary shadowing of the recombinant truncated collagen VII molecules was performed as published previously (35).
Limited Proteolysis-For assessment of the domain structure of the recombinant truncated procollagen VII, 100 g of protein (concentration: 1 mg/ml) was subjected to treatment with collagenase or pepsin as described (48). For deglycosylation, the protein was dialyzed against 20 mM sodium phosphate, pH 7.2, 20 mM EDTA, 0.1% (v/v) Triton X-100 and subsequently reduced with 10 l of ␤-mercaptoethanol for 10 min at 60°C prior to digestion with 5 units of N-glycosidase F (Roche Diagnostics) for 18 h at 37°C.
Digestion with BMP-1-like Enzymes-Recombinant human BMP-1, mTLL-1, and mTLL-2 were prepared as described (18,36); mTLL-1 and mTLL-2 carried a FLAG tag. The digestions with 400 ng of enzyme were performed in 100 -200 l of a buffer containing 50 mM Tris-HCl, pH 8, 150 mM NaCl, 5 mM CaCl 2 , and 1 mM AEBSF for 24 h at 37°C, unless otherwise stated. When recombinant truncated procollagen was used as a substrate, 5 g of the protein were incubated with the respective protease in the presence of 1 mM AEBSF (Merck), 10 g/ml soybean trypsin inhibitor (Sigma), and 10 g/ml leupeptin (Fluka, Buchs, Switzerland). Reactions were stopped by precipitation of proteins by trichloroacetic acid in the presence of deoxycholic acid, and samples were analyzed by SDS-PAGE and immunoblotting. When procollagen VII from cell extracts was used as a substrate, the extracts were concentrated by precipitation with 176 mg/ml ammonium sulfate, and the protein precipitate was harvested by centrifugation for 1 h at 4°C and 13,000 rpm. The protein pellet was resuspended in 400 l/75 cm 2 of the above assay buffer and dialyzed against the same buffer overnight. After dialysis, protease inhibitors were added, and the sample was incubated in the presence of BMP-1 as described above. Protein precipitation and analysis were carried out as mentioned above.
N-terminal Sequencing-After cleavage of the truncated procollagen VII with BMP-1, mTLL-1, or mTLL-2, the cleavage products were separated by SDS-PAGE and transferred to high capacity proteinbinding membrane (Sequi-Blot polyvinylidene difluoride membrane, Bio-Rad). After staining with Coomassie Brilliant Blue, the band cor-responding to the C-propeptide was excised and subjected to N-terminal microsequencing using Edman degradation (Toplab, Martinsried, Germany, and Microsequencing Facility of the Massachusetts General Hospital, Boston).
Analysis of BMP-1 Activity-Media were obtained from primary human keratinocytes and fibroblasts as described above. 94 l of medium were each incubated with 1 g of L-Pro-14 C-human procollagen I for 8 h at 37°C in the presence of 50 mM Tris-HCl, pH 7.5, 5 mM CaCl 2 , 0.02% (w/v) dextran sulfate, 0.01% (w/v) Brij-35, and 0.02% (w/v) sodium azide (49). Incubation of radioactively labeled procollagen I in the absence of cell culture medium served as a negative control, and incubation in the presence of recombinant human BMP-1 served as a positive control, respectively. Samples were separated by 7.5% SDS-PAGE, followed by blotting and exposure to film.
Immunohistochemistry-Collagen VII was detected using the polyclonal antibody NC1-F3 that recognizes the NC-1 domain of collagen VII. 3 The monoclonal antibody "2H1" directed against mTLL-1 was a kind gift from Paul Findell and Suzanne Spong, FibroGen Inc., South San Francisco, CA. The polyclonal antibodies "AB81032" directed against the N terminus of BMP-1/mTLD and "AB81030" against the C terminus of mTLD were purchased from Chemicon, Temecula, CA. The fluorescein isothiocyanate-labeled anti-rabbit/anti-mouse antibodies were from Dako (Glostrup, Denmark). Cells were cultured on coverslips in the presence of 50 g/ml L-ascorbic acid for 48 h, permeabilized, and fixed with absolute methanol at Ϫ20°C. For immunofluorescence staining, cells and tissue cryosections were incubated at room temperature with the first antibody overnight and the second antibody for 1 h. Preparations were mounted with Mowiol (Hoechst, Frankfurt am Main, Germany) and examined using a fluorescence microscope (Axioskop 2, equipped with either an MC 80 DX camera or an AxioCam HRc digital camera, Carl Zeiss, Oberkochen, Germany) at a wavelength of 488 nm. For immunoperoxidase staining, the specimens were incubated with the primary antibody in phosphate-buffered saline containing 1% (w/v) bovine serum albumin overnight at room temperature followed by incubation with the peroxidase-conjugated secondary antibody (EnVision, DAKO, Glostrup, Denmark) for 30 min at room temperature. Peroxidase activity was visualized using chromogenic 3-amino-9-ethylcarbazole (Sigma) in 50 mM acetic acid/sodium acetate, pH 4.7, in the presence of H 2 O 2 at 37°C for 10 min. Cells were counterstained with MayerЈs hemalum solution (Merck) and washed with water. The slides were mounted with Gurr Aquamount (BDH Laboratory Supplies, Poole, UK) and analyzed by light microscopy.

Expression and Characterization of Recombinant Truncated
Procollagen VII-Preliminary experiments with BMP-1 purified from osteosarcoma cell lines indicated that BMP-1 was able to cleave procollagen VII isolated from cultured human keratinocytes. However, only minute amounts of the released C-propeptide were retrieved after the cleavage, impeding the determination of the cleavage site. Therefore, a recombinant truncated procollagen VII was produced for characterization of the BMP-1 processing of procollagen VII and for determination of the cleaved peptide bond. The cDNA construct encoded a fusion product of procollagen VII amino acids 1042-1284 to amino acids 2684 -2944, resulting in the deletion of the Nterminal portion of the NC-1 domain and of the central portion of the triple helical domain (Fig. 1, A and B). However, the von Willebrand factor domain and the cysteine-rich sequences within the NC-1 domain, the N-and C-terminal parts of the triple helix, and the entire NC-2 domain were retained. The construct was thus distinctly smaller than the type VII minicollagen produced by Chen et al. (50) that contained the fulllength NC-1 domain. Truncated procollagen VII was secreted into the conditioned medium of mass cultures of confluent 293-EBNA cells at a concentration of ϳ0.5 mg/liter and was purified by affinity chromatography on nickel-Sepharose. Immunoblotting with domain-specific collagen VII antibodies showed that the recombinant truncated procollagen VII contained the NC-1, triple-helical, and NC-2 domains (Fig. 1C). SDS-PAGE under non-reducing and reducing conditions indicated that the protein was trimeric containing intermolecular disulfide bonds (Fig. 2A). The trimer migrated as a double band leading to the assumption that it consisted of two forms with different intermolecular disulfide bonds and conformations. Both reduced and unreduced monomers migrated as a single band indicating that the double band of the trimer was not due to proteolytic degradation. The unreduced monomer had a smaller apparent molecular weight than the reduced monomer, suggesting also intramolecular disulfide bonding. Incubation with highly purified bacterial collagenase digested the triple helix, yielding the N-and C-terminal globular domains (Fig.  2B), whereas treatment with pepsin removed these domains thus leaving the central helical domain (Fig. 2C). The primary structure predicts one glycosylation site within the truncated NC-1 domain. Incubation of the protein with N-glycosidase F leads to a slight reduction of the molecular weight indicating that the protein is glycosylated (Fig. 2D). These experiments showed that the recombinant protein had the predicted structure and was suitable for use as a substrate for BMP-1.
BMP-1, mTLL-1, and mTLL-2 Process Procollagen VII in Vitro-The truncated procollagen VII was efficiently cleaved by recombinant BMP-1 (Fig. 3A). The C-propeptide was subjected to N-terminal microsequencing, which yielded a major DTAGS sequence. This showed that BMP-1 cleaved the peptide bond between Ala 2821 and Asp 2822 , corresponding to the predicted BMP-1 consensus cleavage site SYAA2DTAG. A second minor sequence obtained was -LHAVP. This sequence is located within the NC-2 domain, six amino acid residues away from the C terminus of the P 1 Ј consensus cleavage site and is likely to represent secondary degradation. The truncated procollagen VII was also subjected to digestion by recombinant human mTLL-1 and mTLL-2, two metalloproteases related to BMP-1 that share at least some substrates with BMP-1. Both enzymes cleaved the truncated procollagen VII, and microsequencing demonstrated that the cleavage occurred at the same Ala 2821 -Asp 2822 bond as with BMP-1. However, cleavage by mTLL-2 seemed to be less efficient because under the same conditions only about half of the substrate was converted. The digestion of truncated procollagen VII by BMP-1 was inhibited by metal chelators, such as 10 mM EDTA and 1 mM o-phenanthroline, or by 5 mM 2,2Ј-bipyridine, 1 mM 1,4-dithio-DL-threitol, and 1 mM Zn 2ϩ but not by serine or cysteine proteinase inhibitors, such as 10 g/ml leupeptin, 1 mM phenylmethylsulfonyl fluoride, 1 mM AEBSF, 2 mM N-ethylmaleimide, or 10 g/ml soybean trypsin inhibitor. Heat denaturation of the procollagen did not inhibit cleavage (not shown).
Rotary shadowing of the recombinant truncated procollagen VII demonstrated the presence of rod-like molecules with small globules at both ends. The measured length of the rod, 35 nm, corresponded to the predicted size of 36 nm calculated from the length of the triple-helical domain of the authentic full-length type VII collagen (27). Digestion of the procollagen with BMP-1 resulted in the removal of one of the globules (Fig. 3B).
To verify the cleavage observed with the recombinant protein, full-length authentic procollagen VII was partially purified from human keratinocyte cultures and used as a substrate for recombinant BMP-1. The reaction resulted in the conversion of the 320-kDa pro-␣1(VII) chains to 290-kDa ␣1(VII) chains, which co-migrated with mature ␣1(VII) chains from normal human dermis. The processing was inhibited in the presence of 25 mM EDTA (Fig. 4).
Procollagen with a Deletion of the BMP-1 Consensus Sequence Is Not Processed-Mutated procollagen VII was isolated from keratinocytes of two DEB patients. Both carried the deletion mutation 8523del14 on one allele of the COL7A1 gene (32). The mutation leads to in-frame skipping of exon 115 that encodes a segment of 29 amino acids containing the putative BMP-1 consensus sequence. In addition, the deletion mutation results in the amino acid substitution E2843Q. On the other allele, patient 1 carried the nonsense mutation A425G and was thus functionally homozygous for the deletion, i.e. synthesized only homotrimers consisting of deleted pro-␣1(VII) chains. Patient 2 had a glycine substitution mutation G2009R on the other allele and was thus compound heterozygous for two different mutations. The prediction was that the keratinocytes of patient 2 synthesized mixed trimers containing deleted pro-␣1(VII) chains and ␣1(VII) chains with a glycine substitution in the central triple-helical domain. Collagen VII was extracted from skin samples of both patients and analyzed by immunoblotting. Only procollagen VII was found in the skin of patient 1, supporting the prediction that the deletion inhibited processing by BMP-1 (Fig. 5A). In the skin of patient 2, procollagen VII was partially processed (Fig. 5A), indicating that pro-␣1(VII) chains with the glycine substitution were cleaved but not the pro-␣1(VII) chains containing the deletion. The inability of BMP-1 to cleave the mutated procollagen VII of patient 1 was confirmed in vitro. Procollagen VII was extracted from keratinocytes of patient 1 and digested with BMP-1. In contrast to normal procollagen VII, no cleavage of the mutated molecules was observed (Fig. 5B).
Cutaneous Cells Express Enzymes of the BMP-1 Family-Immunohistochemical staining with a polyclonal antibody recognizing BMP-1 and its alternatively spliced form mTLD showed a positive signal in cultured keratinocytes and fibroblasts (Fig. 6). In the skin, both mesenchymal and epidermal cells produced a positive signal, with the most prominent staining in basal keratinocytes. The same result was obtained using an antibody specific for mTLD (data not shown). Immunofluorescence staining with a monoclonal antibody directed against mTLL-1 also gave a positive signal in cultured keratinocytes and fibroblasts. However, in the skin, mTLL-1 was detected only in the suprabasal epidermis. This suggests that in vivo, BMP-1/mTLD but not mTLL-1 is the major protease involved in procollagen processing in normal skin. To differentiate between latent pro-forms and mature, biologically active enzymes, culture media of human keratinocytes and fibroblasts were tested for procollagen C-proteinase activity using 14 Clabeled procollagen I as a substrate. It was partially cleaved to pNcollagen and mature collagen I during incubation with fibroblast medium, but no processing was detected with keratinocyte medium (Fig. 7). This suggests that both keratinocytes and fibroblasts produce BMP-1/mTLD but that keratinocytes secrete mainly the latent pro-forms, whereas mesenchymal cells produce most of the enzymatically active enzymes. This is consistent with previous observations (51) using immunoblot analysis that cultured fibrogenic cells secrete mostly mature, active BMP-1/mTLD, whereas keratinocytes produce predominantly the unprocessed precursor.
Procollagen VII Is Processed in BMP-1-deficient Mice-Because procollagen VII can be processed by three enzymes, BMP-1, mTLL-1, and mTLL-2, in vitro, we investigated the processing in the skin of Bmp1 Ϫ/Ϫ mouse embryos (18 dpc). Previous data (41) had shown that homozygous Bmp1 null mice were perinatally lethal, with defects in procollagen I processing and collagen fibrillogenesis, but with residual procollagen Cproteinase activity. In concert with those observations, immunofluorescence staining with collagen VII antibodies produced FIG. 4. Authentic full-length human procollagen VII is processed by BMP-1. Human procollagen VII isolated from normal keratinocytes was digested with BMP-1 and immunoblotted (5% SDS-PAGE) with antibody NC2-10. Lane 1, negative control kept at Ϫ20°C. Lane 2, sample incubated with BMP-1. Lane 3, sample incubated with BMP-1 in the presence of 25 mM EDTA. Lane 4, negative control incubated without BMP-1. Lane 5, mature collagen VII isolated from normal human skin. pro, 320-kDa human procollagen VII; mat., 290-kDa mature human collagen VII.

FIG. 5. Mutant procollagen VII with deletion of the BMP-1 consensus sequence is not processed.
A, immunoblot of skin extracts from two patients with a deletion of the BMP-1 cleavage site. Patient 1 is functionally homozygous for the deletion and has only deleted pro-␣1(VII) chains. Patient 2 is compound heterozygous for the deletion and a missense mutation and has a mixture of deleted pro-␣1(VII) chains and pro-␣1(VII) chains with a glycine substitution within the triple helix. Lanes 1 and 6, procollagen VII (pro) from normal human keratinocytes. Lanes 2 and 5, mature (mat.) collagen VII from normal human skin. Lane 3, collagen VII from the skin of patient 1, only a procollagen band is seen. Lane 4, collagen VII from the skin of patient 2, both procollagen and mature collagen are found. The immunoblot on 5% SDS-PAGE was detected with antibody NC2-10. B, procollagen VII isolated from cultured keratinocytes of patient 1 is not processed by recombinant BMP-1 in vitro. Lane 1, sample incubated without enzyme. a strong reaction along the dermal-epidermal junction in both wild type and Bmp1 Ϫ/Ϫ mouse embryo skin (Fig. 8A), indicating that collagen VII content was not reduced in Bmp1 Ϫ/Ϫ mouse skin. Immunoblotting of dermis extracts of Bmp1 Ϫ/Ϫ mice showed that procollagen VII was processed similarly to wild type mice (Fig. 8B). These data suggest that mTLL-1 and/or mTLL-2 can fully substitute for BMP-1/mTLD in procollagen VII processing in vivo. DISCUSSION Here we show that BMP-1-like proteinases cleave procollagen VII to mature anchoring fibril collagen in the skin. Cleavage of procollagen VII by BMP-1 is similar to the cleavage of procollagens I-III because a C-propeptide is efficiently removed from the procollagen molecule. Noteworthy, the cleavage of the C-propeptide of procollagen V requires furin (5-7). Other BMP-1 substrates, such as probiglycan (8), prolysyl oxidase (21,22), chordin (18), and the ␣3 and ␥2 chains of laminin 5 (9,10,36) have different molecular structures and functions. Thus, BMP-1-like proteinases control a variety of biological functions, ranging from polymerization of different tissue-specific collagen fibrils (23,24), proteoglycan maturation, or crosslinking of collagens and elastin to dorsal-ventral patterning during embryogenesis and to dermal-epidermal cohesion (26).
The importance of the above biological functions is alluded to by the redundancy of the enzymes. For example, as shown here, three members of the astacin family, BMP-1, mTLL-1, and mTLL-2 have the ability to cleave procollagen VII at the same specific site in vitro. Genetic, immunohistological, and activity assays demonstrated that at least BMP-1 and mTLL-1 are produced and active in the skin, whereas the lack of mTLL-2specific antibodies leaves open the question of whether this enzyme might also be expressed at detectable levels in normal skin. To address the issue of whether proteases other than BMP-1 or mTLL-1 are able to cleave procollagen VII, we attempted analysis of procollagen VII processing in the skin of embryos lacking both enzymes. Such embryos are doubly homozygous null for the Bmp1 gene, which encodes both BMP-1 and mTLD, and for the Tll1 gene, which encodes mTLL-1, and were produced by intercrossing heterozygotes from Bmp1 described previously (41) and Tll1 knockout mouse lines (52). However, the doubly null embryos die during embryonic development at 14 -15 dpc, at which stage the skin basement membrane zone is undeveloped, extremely fragile, and not amenable to immunohistological or protein chemical analysis. Thus, although it appears that under physiological conditions BMP-1 and mTLL-1 both appear capable of processing procollagen VII in the skin, it remains to be determined whether these two enzymes are capable of substituting for each other, and whether mTLL-2 may substitute for either or both of the other two related enzymes in vivo. Future experiments should be directed toward resolving such issues as well as assessing possible other functions of these proteinases apart from processing of extracellular matrix proteins.
Early biochemical characterization of procollagen VII demonstrated efficient cleavage of the C-propeptide and predicted that the removal is important for polymerization of the anchoring fibrils (53,54). Genetic evidence supports this notion. In a subset of recessive DEB, a human blistering skin disease, the BMP-1 consensus motif is deleted, and procollagen VII accumulates in the skin. As illustrated by patient 1, who is functionally homozygous (hemizygous) for the deletion of a 29amino acid segment containing the BMP-1 cleavage site, the mutated anchoring fibrils are ultrastructurally abnormal and functionally deficient, clearly demonstrating the significance of adequate procollagen-to-collagen VII processing for adhesion of the epidermis and the dermis.
The dermal-epidermal junction is a large epithelial-mesenchymal interface that regulates a number of skin functions. It consists of structurally complex multiprotein aggregates designed to provide firm adhesion on the one hand between the basal epidermal keratinocytes and the basement membrane and on the other hand between the basement membrane and the papillary dermis. Besides interacting with each other as binding ligands to provide adhesion, the protein components of the junction also mediate cell adhesion, migration, and signal transduction of messages from the extracellular matrix to the keratinocytes. The regulation of these processes occurs at several levels. Post-translational proteolytic cleavage of several of the proteins modifies their functions and is, therefore, an important regulatory mechanism. For example, the shedding of the transmembrane collagen XVII from keratinocyte surfaces (48) or cleavage of laminin 5 arms influences cell adhesion and migration (10,38). Additionally, as shown in the present study, extracellular processing of procollagen VII to collagen is a prerequisite for correct polymerization of the anchoring fibrils. Other proteolytically processed components of the dermal-epidermal junction include ␣ 6 integrin (55) and dystroglycan (56). However, the functional consequences of their cleavage remain elusive.
The laminin 5 precursor and procollagen VII are synthesized and secreted by epidermal keratinocytes. In contrast, active BMP-1-like enzymes seem to originate mainly from mesenchymal fibroblasts (57,58). Accordingly, we could not demonstrate BMP-1-like activity in keratinocyte culture medium, whereas we confirmed the finding that BMP-1/mTLD and maybe other proteases with similar activity are active in culture media of skin fibroblasts. In agreement with the observation of Amano et al. (9) that antibodies to BMP-1 stained basal epidermal keratinocytes in calf skin, we found a positive signal for BMP-1/ mTLD in cultured human keratinocytes in vitro. However, in light of the activity assays, it seems feasible that keratinocytes secrete BMP-1/mTLD predominantly in the latent zymogen forms. This assumption is in agreement with the previous data that show that transforming growth factor-␤1-stimulated keratinocytes produce mainly pro-forms of BMP-1 and mTLD in the presence of a low calcium concentration (0.15 mM Ca 2ϩ ) (51). Pro-BMP-1/mTLD possesses a consensus sequence for processing by furin-type proprotein convertases. Furin itself is known to need a low pH and Ca 2ϩ for activation (59). The primary keratinocytes used in our experiments were maintained also at a low calcium concentration (Ͻ0.1 mM Ca 2ϩ ) to keep the cells at a low state of differentiation similar to the situation found for the basal keratinocytes in the skin. In the presence of higher calcium concentrations (1.5 mM), the keratinocytes start to differentiate, i.e. they form several cell layers and start to cornify. That BMP-1 seems to be activated in the presence of higher calcium concentrations is in agreement with the previous observation that the ␥2 chain of laminin 5 is processed in keratinocyte culture supernatants only in the presence of at least 1 mM Ca 2ϩ (9). The epithelial substratemesenchymal enzyme constellation presents novel ways of regulating epithelial-mesenchymal interactions, such as the assembly of the dermal epidermal junction, cell adhesion, and migration during development and reparative processes.